Re: NFS server panic on OpenBSD diskless client shutdown

[David Holland <dholland-current%netbsd.org@localhost>]

On Thu, Aug 02, 2012 at 10:57:28PM -0500, John D. Baker wrote:
 > VOP_ABORTOP(0,da64ea08,c1e99b00,d,0,da64ea24,da64eb68,da64ea34,da64ea3c,0) 
 > at VOP_ABORTOP+0x1a
 > nfsrv_rename(c5357cf8,c1e99b00,c2246800,da64eb68,0,0,0,da64ec30,0,1) at 
 > nfsrv_rename+0x749

so, it's calling VOP_ABORTOP on a null vnode.

Unfortunately, the logic in rename is pieced together from old
spaghetti, so figuring out which of the many VOP_ABORTOP calls therein
it's dying on will be difficult without instrumenting them with
printfs, or building a -g kernel and using the debug info to get a
source line number for the offset to call site.

I don't see anything immediately and obviously wrong with any of them
but such checking really needs to be done when I'm more awake to
produce useful results.

-- 
David A. Holland
dholland%netbsd.org@localhost